A computational model for automated extraction of structural schemas from simple narrative plots

Abstract

Building computational systems capable of creating and interpreting narrative content has been an objective of Artificial Intelligence since its beginnings. Its development, however, has been blocked by what is commonly known as the knowledge acquisition bottleneck, which does not permit story processing in the large. In this dissertation, a computational system that tries to take one step forward in the target of making it possible to process narrative content on a larger scale is presented. Two stages of research towards this goal are detailed. A semantic approach to narrative processing not yielding satisfying results is first explained. Then, a different system modelling a focus shifting towards a structural management that provided better results is shown. The current state of the art is studied in detail. Empirical validation has been carried out to prove to the possible extent the proposed hypothesis (the plausibility of structural processing for narrative content). Discussion about the most important aspects and design decisions is also included, and possible future lines of investigation are also exposed.